Post

Gift link, and some reactions herein. www.nytimes.com/2026/09/25/o...

The overall conclusion – that current learners are suffering because we don't know yet how to incorporate this technology into training – strikes me as plausible, although I am increasingly skeptical that we're going to reach a "tech is mature and we know how to handle it" equilibrium anytime soon.

This piece and many others like it express a level of hubris – that I'm an experienced clinician and therefore the way I wield this technology is Correct – that I find more than a little concerning. I'm worried about everyone in clinical medicine.

To wit, for example

Bedard mentions but doesn't really go on to engage with what I think is an incredibly important point here – talking to colleagues about challenging cases (and "challenging" evolves over years of training) is one of the most important aspects of medical training and practice.

I've spoken for years about how removing the friction from contacting colleagues has degraded medical practice. In residency, I carried a pager – anyone could page me at any time for any reason, but it carried a social cost, and the friction induced by that cost led to more productive interactions.

Since then, various "instant messaging" platforms have proliferated in hospitals, and friction is at an all time low. Messages are often sent reflexively because it's easier to do so than to check the patient's chart, look at a guideline, or walk to the bedside. pubmed.ncbi.nlm.nih.gov/39033420/

Social friction is good. When you know you are demanding someone's time and incurring their judgment, you're more likely to work through a problem independently to the best of your abilities. This preparatory work makes the ultimate consultation more valuable to the patient and the practitioner.

It also lays the groundwork for higher quality social interactions between practitioners. I was trained – and train learners now – in a medical educational model where learners at every level collaborate.

When I was a senior resident and an intern came to me to ask a question – having worked through the problem themselves but reaching the limits of their abilities – the interaction offered the opportunity to answer the question, but also for me to assess the learner's skills and knowledge,

to provide other relevant teaching (because the learner didn't know enough to know the question they should have been asking), and – and this is really important – to forge a bond based on mutual respect. Having the humility and diligence to ask for help is an incredibly powerful social signal.

We talk a lot about how technology took physicians "away from the bedside," but as an ICU physician – who still spends an awful lot of time at the bedside speaking to my patients (or their conscious family members) – I worry more about technology that is driving physicians away from each other.

Learners I've trained know one of my hobby horses is "stuff you just have to memorize" (for internists: opioid, corticosteroid, and diuretic conversions). Until acute care is delivered through autonomous closed loop systems – which I think is further off than anyone realizes – this will remain true.

We've reached the grab portion of this thread so these are just some observations that didn't fit into the screed above:

I've grown used to seeing this claim in every piece about OE, and it comes directly from the company itself, but it is facially untrue, as I've posted before. OE is as cagey about their model as any for-profit enterprise, but there is plenty of non-medical content in the underlying training corpora.

This is one of the reasons I personally have not incorporated OE into my practice. The vast majority of practicing physicians are either not competent to or do not have the time to individually vet every piece of primary literature supporting a guideline or treatment decision.

I regret that these technologies – which are powerful and could certainly be incredibly useful – have entered medicine in the way they have: unexamined, without rigorous patient-centered evaluation, mired in proprietary claims, regulatory capture, and credulous adopters.